🎧 Prefer to listen?
Only 2% of consumers are actually paying for AI. That single stat, dropped on TechCrunch’s Equity podcast, explains more about this moment than the White House rebranding AI as “super intelligence” or OpenAI giving its agent a bubbly cartoon avatar called Dots. Governments are renaming the technology and consumer apps are getting cuter — because consumers, it turns out, are not the ones paying. I read that stat, looked at my own AI side project’s numbers, and rebuilt it around business customers in an afternoon. Here’s exactly what I did, so you can do a faster version in 20 minutes.
The broader picture first. Last week the White House got nearly every major tech CEO in one room — Zuckerberg, Bezos, Musk, Amodei — to sign an AI safety pledge (which, embarrassingly, misspelled “United States”), and an executive order officially inaugurated the “era of super intelligence.” Meanwhile Meta and OpenAI are slapping friendlier faces on their consumer products. All polish, same problem underneath: the money in AI isn’t coming from people like you and me clicking around a chatbot. It’s coming from businesses paying for outcomes. You can see the same squeeze in what people actually pay for in AI subscriptions.
What I actually did
My side project was a classic consumer AI toy: upload a thing, get a thing back, share it with friends. Free users loved it. Revenue was a rounding error.
I pulled the last 90 days of usage and sorted users by behavior instead of signup source. Two clusters emerged. One was hobbyists who tried it once and left. The other was a small group of people using it weekly for work — a recruiter screening candidates, a real estate agent drafting listing descriptions, an operations manager summarizing vendor emails.
Those weekly users were 4% of signups but 100% of the value. Every one of them had something in common: they weren’t using my tool for fun. They were using it to skip work they hated. That’s a business customer wearing a consumer’s clothes — and I’d been charging them consumer prices.
Do this yourself
You don’t need my data. You need your own, and 20 minutes:
- Export your user list from whatever you use — Stripe, Gumroad, Notion, an Airtable, even a Google Form. Anything with names and usage counts works.
- Mark each user as “play” or “work.” You can usually tell from names (personal Gmail vs a company domain), usage times (weekends vs weekday mornings), and what they ask for in support messages.
- Rewrite your headline for the “work” cluster. Change “Make fun X in seconds” to the job outcome: “Cut your weekly reporting from 3 hours to 10 minutes.” Same tool, business framing.
- Add one business proof element. A single sentence — “Used by teams at 40+ small businesses” — or one concrete before/after number. Business buyers look for proof; consumers look for fun.
- Email three “work” users directly. Ask what they’d pay for a version with priority support and no limits. Their answers become your pricing page.
That’s it. No code, no pivot deck, no relaunch. The same product with a business story in front of it.
What failed
Two honest misses.
I initially changed too much. My first rewrite stripped out every casual feature and repositioned the tool as an “enterprise platform.” My weekly users — the ones I was courting — got confused and one asked if the free version was going away. Business framing is not business pretension. Keep the product; change the story.
The “super intelligence” angle is a trap for marketing. My instinct was to ride the government’s new vocabulary — put “super intelligence” in the headline. It tested terribly. Business buyers don’t buy the era; they buy the outcome. The rebrand is a story about power and policy, not a feature list. If you’re curious where that policy story is heading, the AI landscape after the regulatory crackdown covers the structural side.
Who should skip this
If you’re building for fun and have zero interest in revenue, ignore all of it — consumer delight is a fine goal.
Skip this too if you’re deep in a consumer product with real network effects, like a community or marketplace. The 2% stat describes tools; marketplaces play different economics.
But if you have an AI tool with a handful of power users who use it for work, you’re sitting on the 98%’s opposite: the small group that pays. Don’t wait for a pivot moment — agentic AI is becoming an employee category whether you reposition or not, and the businesses buying that future need tools exactly like the one you already built.
The takeaway
The loudest AI news this month was cosmetic — new names in Washington, new avatars in San Francisco. The number that matters for your bottom line is 2%. If your AI project needs to make money, your next 20 minutes are worth more than the next six months of consumer features: find your work users, speak their outcome, and charge accordingly. And if the sheer number of tools and pivots is paralyzing you, the escape hatch from AI tool overwhelm is a good next read.